[cairo] Success with Cairo on Win9x

[Brian Palmer]

> > Yes, they work just fine. This is the code I used to create a FreeType
> font:
> > cairo_ft_font_face_create_for_ft_face(*((FT_Face*)m_pFreeTypeFace), 0);
> >
> > Then I create a scaled font from that and call cairo_text_path() on a
> string
> > and then stroke the path.
> 
> OK, that could work.

It does work. Without it, the text rendering on Win9x would not happen, but
it does happen.

> > believe I also had to comment out the _cairo_ft_scaled_font_create_toy()
> > function and NULL it's pointer in the backend because that function uses
> > fontconfig stuff extensively.
> 
> But that's a problem. It would mean that cairo_show_text and
> cairo_text_path would not be functional at all.

I don't know about cairo_show_text(), but cairo_text_path() seems to
function just fine without the fontconfig stuff and without
_cairo_ft_scaled_font_create_toy().

> Now, you might be arguing that that's exactly what you want. But if
> so, what you're saying is that you'd like to be able to compile cairo
> without it's toy text API, and that would definitely be something new,
> (perhaps not without precedent though---we do allow cairo to be
> compiled without PNG support which is quite similar to the toy text
> API in one way).

I wouldn't say that compiling without the toy text API is exactly what I
would want. But if it's the only option, then I would think it should be
there. The optimal solution would be to fix
_cairo_ft_scaled_font_create_toy() to not require the fontconfig stuff, at
least not on Win32. That way we could optionally compile out fontconfig.

What exactly is fontconfig used for anyways? I'm not entirely clear on what
it does. Perhaps it would be fairly easy to just replace it's functionality
on Win32 with something just as good, or better?